sign in sign up
<<
Home , Aira caryophyllea, Ampelodesmos, Amphibromus, Andropogoneae, Anisopogon, Anomochloa

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Keck Graduate Institute

Aliso: A Journal of Systematic and Evolutionary Botany

Volume 23 | Issue 1 Article 27

2007 A Preliminary Phylogenetic Analysis of Grass Subfamily (), with Attention to Structural Features of the Plastid and Nuclear Genomes, Including an Intron Loss in GBSSI Jerrold I Davis Cornell University, Ithaca,

Robert . Soreng Smithsonian Institution, , ..

Follow this and additional works at: http://scholarship.claremont.edu/aliso Part of the Botany Commons, and the Ecology and Evolutionary Biology Commons

Recommended Citation Davis, Jerrold I and Soreng, Robert J. (2007) "A Preliminary Phylogenetic Analysis of the Grass Subfamily Pooideae (Poaceae), with Attention to Structural Features of the Plastid and Nuclear Genomes, Including an Intron Loss in GBSSI," Aliso: A Journal of Systematic and Evolutionary Botany: Vol. 23: Iss. 1, Article 27. Available at: http://scholarship.claremont.edu/aliso/vol23/iss1/27 Aliso 23, pp. 335–348 ᭧ 2007, Rancho Santa Ana Botanic Garden

A PRELIMINARY PHYLOGENETIC ANALYSIS OF THE GRASS SUBFAMILY POOIDEAE (POACEAE), WITH ATTENTION TO STRUCTURAL FEATURES OF THE PLASTID AND NUCLEAR GENOMES, INCLUDING AN INTRON LOSS IN GBSSI

JERROLD IDAVIS1,3 AND ROBERT J. SORENG2 1L. . Bailey Hortorium and Department of Biology, Cornell University, Ithaca, New York 14853, USA; 2Department of Botany and U. S. National Herbarium, National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20013-7012, USA, ([email protected]) 3Corresponding author ([email protected])

ABSTRACT Phylogenetic relationships in the grass family (Poaceae), with specific attention to the internal structure of subfamily Pooideae, are analyzed on the basis of nucleotide sequence variation in plastid- encoded genes (matK, ndhF, ndhH, and rbcL). The resulting phylogenetic hypothesis was examined with attention to the taxonomic distributions of two inversions and an insertion/deletion within ndhF, the absence of intron 10 of the nuclear gene GBSSI (waxy), and positions of the boundaries between the Short Single Copy (SSC) region and the neighboring Inverted Repeat (IR) regions of the plastid genome, relative to the endpoints of ndhF and ndhH, which span these boundaries in some taxa. The PACCAD clade is resolved, and extension of the 3Ј-end of ndhF from the SSC region into the IR region is interpreted as a synapomorphy of this clade. The BEP clade also is resolved, with Ehrhar- toideae placed as the sister of a clade in which Bambusoideae and Pooideae are sister groups. The loss of GBSSI intron 10 is interpreted as a synapomorphy of s.., which includes the traditionally defined tribes Poeae, Aveneae, and Hainardieae, and the results support a novel set of relationships among the tribes of Pooideae, including the placement of Brachypodieae, Bromeae, , and Poeae s.l. within a clade for which a three-nucleotide inversion in ndhF is interpreted as a synapo- morphy, while a six-nucleotide inversion in ndhF marks a clade that includes all sampled members of subtribe Aveninae within Poeae s.l. Key words: GBSSI, intron, matK, ndhF, ndhH, phylogenetics, Poaceae, Pooideae, rbcL, systematics.

INTRODUCTION among the PACCAD clade, Pooideae (sensu GPWG 2001), and two other subfamilies, Ehrhartoideae (formerly Oryz- Phylogenetic Relationships oideae) and Bambusoideae (sensu GPWG 2001), by Clark The grass family (Poaceae) has been a constant subject of et al. (1995), Soreng and Davis (1998), Hilu et al. (1999), attention by systematists, who have evaluated relationships and the GPWG (2001). A ‘‘BEP clade’’ (consisting of Bam- in this group for hundreds of years, on the basis of morpho- busoideae, Ehrhartoideae, and Pooideae; originally the BOP logical, anatomical, and other features (see reviews by Steb- clade; Clark et al. 1995) frequently has been resolved, but bins and Crampton 1961; Soreng and Davis 1998; Grass alternative structures also have been observed. Phylogeny Working Group [GPWG] 2001; and citations Although monophyly of subfamily Pooideae (sensu therein). In recent years, with the development of formal GPWG 2001) has been supported by most of the analyses methods of phylogenetic analysis, several studies have fo- cited above, the relationships resolved among its constituent cused on higher-level relationships in the grasses, using a groups have varied widely. As the present contribution is a variety of structural and molecular characters (.., Kellogg preliminary report of relationships resolved by DNA nucle- and Campbell 1987; Davis and Soreng 1993; Clark et al. otide sequence data, we do not provide a detailed review of 1995; Soreng and Davis 1998; Hilu et al. 1999; Hsiao et al. the various relationships that have been supported by pre- 1999; Mathews et al. 2000; Zhang 2000; GPWG 2001; Du- vious studies, but refer the reader to two analyses that have vall et al. 2007). Most of these analyses provide similar or sampled broadly within this large subfamily, using restric- identical results with respect to many of the major group- tion site data and morphology, and that discuss the problems ings. For example, one group that now appears to be well attendant to the phylogeny of this group (Soreng and Davis established is the ‘‘PACCAD clade’’ (consisting of Panicoi- 1998, 2000). One of the principal findings of the latter paper deae, , , Centothecoideae, Ar- was that neither Poeae nor Aveneae are monophyletic as istidoideae, and ; originally the PACC clade; traditionally circumscribed (cf. Clayton and Renvoize 1986). Davis and Soreng 1993), which includes more than half of Rather, elements of these two groups are intermixed within all of the family (Clayton and Renvoize 1986). An- a clade that also includes the small tribe Hainardieae. We other group that is usually resolved is subfamily Pooideae, refer to this assemblage as Poeae s.l., and in the present work which as circumscribed by the GPWG (2001) includes about we examine relationships within this clade, within Pooideae, one-third of all grass species (Clayton and Renvoize 1986). and within Poaceae as a whole, with particular attention to However, important differences exist among the results of a series of structural characters of the plastid and nuclear these analyses, including the disparate relationships resolved genomes. 336 Davis and Soreng ALISO

Fig. 1.—Map, not to scale, of a portion of the plastid genome of Triticum aestivum L., adapted from GenBank accessions AB042240 and NC002762. Upper portion of figure depicts genes and intergenic spacer regions near endpoints of the Short Single Copy (SSC) region and adjacent portions of flanking Inverted Repeat (IR) regions, with the central portion of the SSC region excised, and with boundaries between the regions indicated by vertical lines. Genes depicted above the horizontal line, with arrows pointing to the right, are encoded from left to right (i.e., their 5Ј-ends are on the left), and genes depicted below the horizontal line, with arrows pointing to the left, are encoded from right to left. Lengths of genes and intergenic spacers are indicated by numbers of base pairs (bp) above the horizontal line. Lower portion of figure depicts positions of two fragments that were amplified and sequenced to determine the locations of the 5Ј-endpoint of ndhH and the 3Ј-endpoint of ndhF relative to the SSC/IR boundaries (see text).

Inversions and Gene Positions in the Plastid Genome Poaceae, Ecdeiocoleaceae, and Joinvilleaceae, but none of Restionaceae or any other family. The third most inclusive The plastid genomes of grasses exhibit a set of three in- clade, marked by the smallest of the inversions (a few hun- versions that differentiate them from those of most other dred base pairs [bp] in length, encompassing trnT), includes , and analyses of the distributions of these inversions all sampled species of Poaceae and no others. The distri- have contributed to an understanding of relationships among butions of these inversions, in concert with morphological grasses and other families. When the complete nucleotide and nucleotide sequence data (Bremer 2002; Michelangeli et sequence of the plastid genome of Oryza L. was published al. 2003), have provided compelling evidence that Joinvil- (Hiratsuka et al. 1989), it was observed to differ from that leaceae and Ecdeiocoleaceae are the closest living relatives of Nicotiana L. by three inversions in the Large Single Copy (LSC) region. Subsequent analyses (Doyle et al. 1992; Ka- of the grasses. Consequently, elements of these two small tayama and Ogihara 1996; Michelangeli et al. 2003) have families now are widely used as outgroups for phylogenetic indicated that the three inversions demarcate three clades studies of the grasses, and they are employed in that manner that appear to be perfectly internested. Until every species here. has been sampled, the precise status of every taxon with Another example of structural variation among the plastid respect to these inversions cannot be known, but a fairly genomes of grasses involves the locations of two genes, well-defined picture has emerged, as follows. The most in- ndhF and ndhH, relative to the boundaries of the major re- clusive of the three clades, marked by the largest of the gions of this genome (Ogihara et al. 2002). These two genes inversions (28 kilobases [kb] in length), includes all species are positioned near opposite ends of the Short Single Copy of Poaceae, Joinvilleaceae, and Ecdeiocoleaceae, and some (SSC) region, close to the borders with the two Inverted or all elements of Restionaceae (cf. Katayama and Ogihara Repeat (IR) regions that flank it (Fig. 1). The IR regions are 1996; Michelangeli et al. 2003). This group includes four of identical copies of the same DNA sequence in reverse ori- the seven families of sensu Dahlgren et al. (1985), entation (see Palmer 1983; Plunkett and Downie 2000; and but the absence of this inversion from the remaining families citations within), so all genes and intergenic regions that lie (e.g., Flagellariaceae), and possibly from some elements of within them are present as two copies in the plastid genome, Restionaceae, does not imply that Poales as circumscribed oriented in opposite directions relative to the overall struc- by Dahlgren et al. (1985) are non-monophyletic, for these ture of the plastid genome (e.g., rps15 in Fig. 1). Exami- taxa still may be the closest relatives of the clade that is nation of the first three plastid genome sequences of grasses marked by the inversion. However, the potential absence of to be published (Oryza, Triticum L. and L.) indicates this inversion from some species of Restionaceae does sug- that a substantial portion of the 5Ј-end of ndhH extends into gest that this family may not be monophyletic. The next the IR region in Oryza and Triticum, and thus is duplicated most inclusive clade, marked by the second largest of the in these taxa (Fig. 1; also see Ogihara et al. 2002: Fig. 5). inversions (6 kb in length), includes all sampled species of In Zea, a portion of the 3Ј-end of ndhF extends into the IR VOLUME 23 Phylogeny of Poaceae Subfamily Pooideae 337 region, and thus is duplicated in a similar manner, but one Two of the four plastid-encoded genes lie within the LSC nucleotide of ndhH in Zea also extends into the IR. Because region (matK and rbcL), while the other two (ndhF and different ends of the two genes lie near or within the IR ndhH) are principally in the SSC region, though each ex- region in all of these taxa (the 5Ј-end of ndhH and the 3Ј tends into the IR region in some taxa, as noted above. In the -end of ndhF), any portion of ndhH that extends into the IR course of sequencing ndhF and ndhH, the regions adjacent region is read in one direction (from the 5Ј-end within the to the 3Ј-end of the former and the 5Ј-end of the latter, col- IR region towards the SSC region), while any portion of lectively referred to here as the SSC/IR regions, also were ndhF that extends into the IR region is read in the other sequenced (Fig. 1). The first of the two SSC/IR regions, the direction (from the 5Ј-end in the SSC region towards the IR ndhF/rps15 region, is amplified with a primer situated at ca. region). Thus, the solitary nucleotide of ndhH that lies with- nucleotide 80 of rps15, and another at ca. nucleotide 1810 in the IR region in Zea simultaneously encodes the first base of ndhF. The second SSC/IR region, the ndhA/rps15 region, pair of ndhH and an internal site within ndhF (i.e., the first is amplified with the same primer within rps15 that is used site of ndhF that lies within the IR), as read in opposite for the first SSC/IR region, in combination with a primer at directions. In summary, ndhH extends into the IR in all three ca. nucleotide 60 of ndhA. The positions of the boundaries of these grasses, though to varying extents (207 bp in Trit- between the two ends of the SSC region and the adjoining icum, 163 bp in Oryza, and 1 bp in Zea), while ndhF extends IR regions were determined, along with the positions of the into the IR only in Zea. Ogihara et al. (2002) concluded from 3Ј-terminus of ndhF and the 5Ј-terminus of ndhH relative to this similarity between Oryza and Triticum that these two these boundaries, by comparing the sequences of these two taxa are more closely related to each other than either is to regions, starting from the end of each that includes a portion Zea, but in the absence of evidence from additional taxa this of rps15, and reading toward the SSC/IR boundary. For the conclusion is unsupported, for the similarity between Oryza phylogenetic analysis, the presence vs. absence of any por- and Triticum could be plesiomorphic within the smallest tion of ndhF within the IR region was coded as one binary clade that includes all three of these genera. The taxonomic character, and the presence vs. absence of any portion of distribution of the positions of these two genes, relative to ndhH within the IR region as another. the SSC/IR boundaries, is examined in the present study, Most sequences used in the analysis were generated from with attention to their phylogenetic implications. total genomic DNA isolations of vouchered collections, fol- lowing standard PCR and automated cycle-sequencing pro- GBSSI Intron 10 tocols, though some were obtained from GenBank (Table 1). Primers used for amplification and sequencing of rbcL and There are many cases in which intron losses in the three most of ndhF have been published previously (rbcL: Chase principal genomes of plants have provided evidence of phy- et al. 1993; Asmussen and Chase 2001; ndhF: Olmstead and logenetic affinities (e.g., Wallace and Cota 1996; Frugoli et Sweere 1994). Additional primers used for matK and the al. 1998; Itchoda et al. 2002). The phylogenetic results pre- SSC/IR regions (including ndhH and the 5Ј-end of ndhF) sented here are derived from structural and sequence varia- were developed for this project, and will be described in a tion in the plastid genome, but we have recently initiated a forthcoming contribution. Sequences were aligned manually, study of sequence variation within Poeae s.l. in the nuclear- and regions in which alignment was considered ambiguous encoded gene granule-bound starch synthase (GBSSI, or were excluded from analyses. waxy). Variation patterns in GBSSI have proven useful in In addition to the nucleotide sequence data and the posi- analyses at a range of taxonomic levels within the grasses, tions of the endpoints of ndhF and ndhH relative to the SSC/ from the overall phylogenetic structure of the family to that IR boundaries, several additional structural features were among closely related species, including the details of po- scored as characters for the analysis. One of these is a 15- lyploidization events (e.g., Mason-Gamer et al. 1998; nucleotide insertion/deletion (indel) (Clark et al. 1995, 2000; GPWG 2001; Mason-Gamer 2001; Mathews et al. 2002; In- also discussed as character 53 by GPWG 2001) correspond- gram and Doyle 2003). In the present work we report a phy- ing to sites 1704–1718 in the ndhF sequence of Triticum logenetically informative loss of GBSSI intron 10 within (which is undeleted for these sites) in GenBank reference Poeae s.l. accession NC002762, and situated approximately at site 1704 in the ndhF sequence of Oryza (which is deleted for MATERIALS AND METHODS these sites; precise position is uncertain due to ambiguity in alignments) in GenBank reference accession NC001320. Molecular Data Two additional structural variants in ndhF are interpreted Nucleotide sequence variation was examined in matK, as inversions of three and six nucleotides in length, respec- ndhF, ndhH, and rbcL from 106 representative species of tively. The three-nucleotide inversion corresponds to sites Poaceae and one species each from two related families, 1918–1920 in the reference sequence of Triticum. For most Joinvilleaceae and Ecdeiocoleaceae (Table 1). Of the 424 taxa in the sample the sequence of these three nucleotides gene/taxon combinations, all but four sequences are in the is either TAC (as in Triticum) or its reverse complement, working data set. In some cases (e.g., Schrad. GTA. Taxa with sequences other than these two combina- ex Nees; Table 1) the sequences for a given terminal in the tions usually differ from one or the other of these three- analysis represent different species. This use of ‘‘conglom- nucleotide sequences at no more than one of the three sites, erate taxa’’ represents a compromise between the goal of as would be consistent with the occurrence of point muta- consistency in sampling and that of including as many - tions having caused differentiation from one or the other of quences as possible in the analysis. the two principal sequences. The six-nucleotide inversion 338 Davis and Soreng ALISO

Table 1. Taxa sampled for plastid genome sequence variation, voucher collection information (herbarium acronyms as in Holmgren et al. 1990), and GenBank nucleotide sequence accession numbers. Asterisks identify taxa for which all sequences were obtained from GenBank; these names are accompanied in the left column by accession numbers of those sequences. Non-asterisked taxon names are those for which the authors generated one or more sequences, though additional sequences may have been obtained for these taxa from GenBank, as indicated by accession numbers with these names in the left column. GenBank accession numbers in the right column identify sequences generated by the authors from the specified plant accessions. GenBank numbers are absent for sequences generated by the authors but not yet released.

Voucher information for sequences Taxon generated by authors

Achnatherum occidentalis (Thurb. ex S. Watson) Bark- Soreng 7418 (US) worth subsp. pubescens (Vasey) Barkworth tenerrima Trin. Soreng 3734 (BH) caryophyllea L. Soreng 5953b (US) magellanicus Lam. Soreng 3514 (BH) mauritanica (Poir.) . Durand & Schinz Soreng & Soreng 4029 (BH) scabrivalvis (Trin.) Swallen Soreng 7013 (US) strictus . Br. [rbcL U88403.1]; *A. cari- Linder 5634 (BOL) cinus . Muell. [matK AF31274.1] avenaceus R. Br. Linder 5590 (BOL) marantoidea Brongn. [matK AF164381.1; Davis 753 (BH) rbcL AF021875.1] odoratum L. Soreng 4292 (BH) latifolia (R. Br.) Griseb. Soreng 6016 (US) donax L. [matK AF164408.1; rbcL U31360.1] Crisp 278 (CANB) sativa L. ‘ASTRO’ Davis 759 (BH) Avenella flexuosa (L.) Drejer. Soreng 7305b (US) multiplex (Lour.) Raeusch. ex Schult. & Davis 770 (BH) Schult. f. syzigachne (Steud.) Fernald Soreng 3513 (BH) variegata (Lam.) Kergue´len Grown from USDA Plant Intro. Sta. 253455 (no voucher) erectum (Schreb.) P. Beauv. Soreng 3427a (BH) pinnatum (L.) P. Beauv. Davis 760 (BH); grown from USDA Plant Intro. Sta. 440170 [ndhF AY622312.1; rbcL AY632361.1] . sylvaticum (Huds.) P. Beauv. Soreng 5923b (US) minor L. Davis 761 (BH); grown from USDA Plant Intro. Sta. 378653 inermis Leyss. [matK AF164398.1; rbcL Davis 762 (BH); grown from USDA Plant Intro. Sta. 314071 Z49836.1] B. suksodorfii Vasey Soreng 7412 (US) canadensis Michx. Soreng 7414 (US) Calotheca brizoides (Lam.) Desv. Soreng 7014 (US) aquatica (L.) P. Beauv. Soreng 3861 (BH) subaristatum (Lam.) Desv. Soreng 7020 (US) Chasmanthium nitidum (Baldwin) H. . Yates; *C. lax- Wipff & Jones 2075 (TAES) um (L.) H. O. Yates [matK AF164414.1] aff. subulata L. G. Clark Peterson & Judziewicz 9499 (US) memphitica (Spreng.) . Richt. Boulos & Cope 17676 (E) cristatus L. Grown from RBG, Kew, seed bank 39006 (K) Dacytlis glomerata subsp. hackelii (Asch. & Graebn.) Soreng 3692 (BH) Cif. & Giacom. californica Bol. Davis 763 (BH); grown from USDA Plant Intro. Sta. 232247 cespitosa (L.) P. Beauv. subsp. cespitosa Soreng 7417 (US) sicula (Jacq.) Dumort. Grown from RBG, Kew, seed bank 17332 (K) obovata (Gleason) Brandenberg Davis 756 (BH) floribunda (Pilg.) Pilg. Peterson et al. 14566 (US) Dupontia fisheri R. Br. Cayouette & Dalpe´ C6779-4 (DAO) brachypodium (P. Candargy) Keng & Keng f. Soreng 5358 (US) Ecdeiocolea monostachya F. Muell. Conran et al. 938 (PERTH, ADU) [ndhF AY622313.1; rbcL AY123235.1] caespitosis C. E. Hubb. Soreng 5900 (US) calycina Sm. Grown from USDA Plant Intro. Sta. 208983 trachycaulus (Link) Gould ex Shinners Soreng 4291b (BH) Do¨ sp. US National Herbarium Greenhouse 153, Soderstrom 2182 (US); or US Na- tinoal Herbarium Greenhouse 286 (no voucher) rubra L. Soreng 7424 (US) ventricosum (Gouan) Schinz & Thell. Grown from RBG, Kew, seed bank 5430 (K) fragilis (L.) P. Beauv. Davis 764 (BH); grown from USDA Plant Intro. Sta. 442496 VOLUME 23 Phylogeny of Poaceae Subfamily Pooideae 339

Table 1. Continued.

Voucher information for sequences Taxon generated by authors

Glyceria grandis S. Watson No voucher [ndhF AY622314.1; rbcL AY632364.1] angustifolia Kunth Peterson & Judziewicz 9527 (US) * marantifolia Franch. [ndhF AF164777.1; All sequences obtained from GenBank rbcL AF164778.1] sagittatum (Aubl.) P. Beauv. Clark & Asimbaya 1472 (ISC) convolutum (C. Presl) Henrard Soreng 3803 (BH) comata (Trin. & Rupr.) Barkworth Soreng 7431 (US) annuus Salzm. ex C. A. Mey. Soreng 3642 (BH) * vulgare L. [matK AB078138.1; ndhF All sequences obtained from GenBank U22003.1; ndhH AJ011848.1; rbcL AY137456.1] Joinvillea gaudichaudiana Brongn. & Gris [matK Davis 751 (BH) AF164380.1, published as J. ascendens Gaudich. ex Brongn. & Gris] Leucopoa kingii (S. Watson) . A. Weber Soreng 3515 (BH) pauciflora (Sw.) P. Beauv. [matK Clark 1297 (ISC) AF164385.1] tibetica Hemsl. Soreng 5487, 5490, 5494 (US) perenne L. Davis 765 (BH); grown from USDA Plant Intro. Sta. 418710 spartum L. Soreng 3698 (BH) cupanii Guss. Davis 766 (BH); grown from USDA Plant Intro. Sta. 383702 [ndhF AY622315.1; rbcL AY632365.1] Merxmuellera macowanii (Stapf) Conert [rbcL Barker 1008 (BOL) U31438.1] . rangei (Pilg.) Conert [rbcL AY640153.1] Barker 960 (GRA) minima (L.) Desv. Devesa 3885 (BH) vernale M. Bieb. Soreng 3748 (BH) Molineriella laevis (Brot.) Rouy Soreng 3613 (BH) caerulea (L.) Moench [matK AF164411.1] No voucher [rbcL AY632367.1] strictus L. Royl & Schiers s. . (1988, B) pulchra (Hitchc.) Barkworth Soreng 7407 (US) N. viridula (Trin.) Barkworth No voucher; grown from USDA Plant Intro. Sta. 387938 latifolia L. [matK AF164386.1] Peterson & Annable 7311 (US) *Oryza nivara Sharma & Shastry [all five genes All sequences obtained from GenBank NC005973.1] *O. sativa L. [all five genes NC001320.1] All sequences obtained from GenBank asperifolia Michx. Soreng 5989 (US) incurva (L.) C. E. Hubb. Grown from RBG, Kew, seed bank 24867 (K) radiciflora Sagot ex Do¨ll [matK AF164387.1] Clark & Zhang 1344 (ISC) [rbcL AY632369.1] globosa Munro ex Benth. Clark 1292 (ISC) [rbcL AY632370.1] latifolius L. [matK AF164388.1; rbcL No voucher AY357724.1] pratense L. Soreng 4293 (BH) miliaceum (L.) Coss. Davis 767 (BH); grown from USDA Plant Intro. Sta. 284115 [ndhF AY622317.1] refractus (A. Gray) Benth. Soreng 3381 (BH) alpina L. Soreng 6115-1 (US) monspeliensis (L.) Desf. No voucher Pseudosasa japonica (Siebold & Zucc. ex Steud.) Ma- Davis 771 (BH) kino ex Nakai distans (Jacq.) Parl. Davis 755 (BH) * ciliata Franch. [rbcL AF164780.1]; *P. olyri- All sequences obtained from GenBank formis (Franch.) Clayton [ndhF AF182345.1] pubescens (Desf.) Tzvelev Soreng 3793 (BH) * officinarum L. [all five genes NC006084.1] All sequences obtained from GenBank purpurascens (Torr.) Swallen Soreng 3348 (BH) dura (L.) P. Beauv. Soreng 3862 (BH) caerulea (L.) Ard. Davis 768 (BH); original collection Scholz s. n. (B) Sinochasea trigyna Keng Soreng 5644 (US) divaricatus (Gouan) Rchb. Soreng 3700 (BH) Sporobolus giganteus Nash Peterson et al. 10008 (US) barbata Desf. Davis 768 (BH); grown from USDA Plant Intro. Sta. 229468 340 Davis and Soreng ALISO

Table 1. Continued.

Voucher information for sequences Taxon generated by authors

Stipagrostis zeyheri (Nees) DeWinter [rbcL U31378.1] Barker 1133 (BOL) Streptochaeta sodiroana Hack.; *S. angustifolia - Peterson & Judziewicz 9525 (US) [ndhF AY622318.1; rbcL AY632372.1] erstr. [matK AF164382.1] Timouria saposhnikovii Roshev. Soreng 5448 (US) pauciflora (J. Presl) G. L. Church Davis 533 (BH) Trikeraia pappiformis (Keng) P. C. Juo & S. L. Lu Soreng 5653 (US) nitens (Guss.) Link Soreng 3701 (BH) canescens Buckley Soreng 3383a (BH) *Triticum aestivum L. [all five genes NC002762.1] All sequences obtained from GenBank paniculata L. [matK AF144607.1] No voucher [rbcL AY632373.1] Vahlodea atropurpurea (Wahlenb.) Fr. ex Hartm. Soreng 6316 (S) microstachys (Nutt.) Munro Soreng 7406 (US) *Zea mays L. [all fives genes NC001666.2] All sequences obtained from GenBank corresponds to sites 1932–1937 in the reference sequence of is unambiguous (amb-), with support for a group regarded Triticum. The sequence of these six nucleotides is GAAAAA as unambiguous only when the length of the group’s sub- or its reverse complement, TTTTTC, in most taxa in the tending branch is greater than zero under all possible char- sample. As with the three-nucleotide inversion, most taxa acter optimizations. The search strategy involved 1000 in- with sequences other than these two combinations differ dividual search initiations, using random taxon entry se- from one or the other of these sequences at no more than quences, with each initiation followed by tree-bisection-re- one of the sites (e.g., the sequence in this region is connection (TBR) swapping with up to 20 shortest trees TAAAAA in Triticum). These two inversions were scored retained and subjected to additional branch swapping, using as binary characters for the analysis, as was the 15-nucleo- the command mult*, preceded by rs 0 and hold/20. Six tide indel described above, and eight additional indels ob- unique most-parsimonious trees were generated by these served among the four genes. In the scoring of the two in- 1000 search initiations, and because all of these trees already version characters, taxa with sequences corresponding ex- had been subjected to complete rounds of branch swapping actly to the most commonly observed sequences described without generating additional trees, no further branch swap- above, which are reverse complements (TAC or GTA for the ping was conducted. three-nucleotide inversion, and GAAAAA or TTTTTC for Support for clades resolved by the cladistic analysis was the six-nucleotide inversion), and taxa differing from these assessed by jackknife analysis (Farris et al. 1996), using four standard sequences at no more than one site, were WinClada vers. 1.00.08 (Nixon 2002) running NONA as a scored as having one state or the other. Taxa with sequences daughter process for tree searches, and employing the same other than these were scored as unknown. character and polytomy settings that were used in the basic One additional structural character, presence/absence of analyses of relationships. The jackknife analysis consisted of GBSSI intron 10, also was encoded as a binary character. 1000 replicates, each replicate consisting of four search ini- Observations relating to this character were obtained from tiations, with up to 20 trees retained during TBR swapping sequences in GenBank and from new sequences generated after each search initiation (hold/20; mult*4), followed by in the course of the present work, which otherwise are not additional TBR swapping of all shortest trees, including included in the present analysis. Like the other structural those generated during this phase of swapping, with up to characters, presence/absence of GBSSI intron 10 was encod- 100 trees retained (hold 100; max*). ed as a binary character and included in the analysis. RESULTS Data Analysis Data Characteristics Nucleotide sequences of the four genes and scores for the structural characters described above were combined into After the removal of ambiguously aligned regions, the to- one matrix that was subjected to cladistic analysis, with all tal number of aligned nucleotide sites in each gene, along characters weighted equally and treated as nonadditive (i.e., with the number of cladistically informative sites, and the the states unordered) during tree searches, and with Joinvil- percentage of aligned sites that are informative, are as fol- lea Gaudich. ex Brongn. & Gris as the outgroup for the lows: ndhF (2129, 594, 28%); ndhH (1206, 275, 23%); matK purpose of rooting. Parsimony searches and a jackknife anal- (1590, 544, 34%); rbcL (1344, 237, 18%). Also, there are ysis were conducted following the removal of cladistically 13 cladistically informative structural characters. Thus, a to- uninformative characters from the data set. Parsimony anal- tal of 1663 informative characters were included in the anal- yses were conducted with the multi-thread version of NONA ysis, with ca. 36% of these representing nucleotide sequence vers. 1.6 (i.e., ‘‘PARANONA,’’ compiled 26 Feb 1998; Go- variation in ndhF, 33% in matK, 17% in ndhH, 14% in rbcL, loboff 1993), using the default polytomy settings, which al- and less than 1% representing genomic structural characters. low polytomies to occur (poly ϭ), and which resolve a clade, Cladistic analysis of the combined matrix of 1663 informa- rather than a polytomy, only when support for the resolution tive characters resolved six most-parsimonious trees of VOLUME 23 Phylogeny of Poaceae Subfamily Pooideae 341 length 6946, with a consistency index (CI) of 0.36 and a pling within the PACCAD clade is minimal, and we do not retention index (RI) of 0.73. Lengths, CIs, and RIs on these discuss relationships within this group further, except to note trees for the informative characters of the five data partitions that a clade consisting of Centothecoideae and are as follows: ndhF (2626–2627, 0.36, 0.73); ndhH (1261– (the latter including Gynerium Willd. ex P. Beauv.; see be- 1266, 0.31, 0.70); matK (1987–1992, 0.42, 0.77); rbcL low) is sister to one that includes the other four subfamilies, (1033–1034, 0.31, 0.71); structural characters (33, 0.51, as in the results of the GPWG (2001). Within the BEP clade, 0.86). Because sequences of all four genes were not avail- Ehrhartoideae are placed as the sister of a clade in which able for all 108 taxa, these numbers partially reflect the Bambusoideae and Pooideae are sister taxa. availability of more sequences for some genes than for oth- ers, but an analysis of the subset of 104 taxa that are com- Inversions and Gene Positions in the Plastid Genome plete for all four genes yielded numbers that differ only neg- ligibly from those that are reported. The sequence in the three-nucleotide inversion region in ndhF is either TAC, or a sequence differing from TAC at Phylogenetic Relationships one site, or ‘‘unknown’’ in all taxa within the clade that consists of Brachypodium P. Beauv. and its sister group (Fig. The consensus of the six most-parsimonious trees is de- 3), with the one exception of Hordeum L. In particular, the picted in Fig. 2 and 3. In these trees the grass family is sequence in both species of Brachypodium is TAC. The prin- resolved as monophyletic. Relationships among groups with- cipal alternative sequence (GTA), or a sequence differing in Poaceae are described here in terms of the GPWG (2001) from GTA at one site, or ‘‘unknown’’ occurs in all other classification, to the extent that the present taxon sampling taxa in the sample that have been scored, including Hordeum duplicates that of the GPWG. Sampling of Pooideae in the and all four representatives of (the sequence is present analysis is substantially more extensive than in the GTA in all five of these taxa). The transformation from GTA GPWG analysis, and several taxa placed by the present anal- to TAC therefore is interpreted as a synapomorphy of the ysis within a clade that is conventionally recognizable as clade that is depicted in Fig. 3 (internode C), and Hordeum Pooideae have seldom or never been included in Pooideae is interpreted as having experienced a reversion to the ple- in previous treatments. Thus, we provisionally designate the siomorphic state. clade that consists of Brachyelytrum P. Beauv. and its sister The sequence of the six-nucleotide inversion region in group (Fig. 2, 3) as Pooideae, and to minimize repetition a ndhF occurs as state TTTTTC in Avena L., Helictotrichon detailed description of that group is presented only in the Besser ex Schult. & Schult. f., Trisetum Pers., Rostraria Discussion. Trin., and Gaudinia P. Beauv. In all other taxa in the sample Of the 12 subfamilies delimited by the GPWG (2001), all the sequence in this region is GAAAAA, or a sequence dif- except , Centothecoideae, and Aristidoideae are fering from this sequence at one nucleotide, or a sequence represented in this analysis by more than one exemplar each, scored ‘‘unknown.’’ The transformation from GAAAAA to and thus subject to testing for monophyly. With the matter TTTTTC therefore is interpreted as a nonhomoplasious syn- of Pooideae temporarily set aside, all but one of the remain- apomorphy of the specified five-taxon clade (Fig. 3, inter- ing eight subfamilies are resolved as monophyletic (Fig. 2, node E). 3). The exception is . The two elements of The deleted form of the 15-nucleotide indel in ndhF oc- this subfamily that are sampled in the present study, repre- curs in Joinvillea, Ecdeiocolea F. Muell., Anomochloa, senting the only two genera in the subfamily, are united as Streptochaeta, Pharus P. Browne, both species of Oryza, and a clade in two of the six trees. In the other four, Strepto- Nardus L. The undeleted form occurs in all other taxa in the chaeta is placed as the sister of a clade that includes all other sample. As previously reported by Clark et al. (1995, 2000) grasses, with Anomochloa Brongn. as the next line to diverge and the GPWG (2001), this character is interpreted most from the clade that includes all remaining grasses. The con- parsimoniously as an insertion in the lineage that includes sensus of these structures (Fig. 2) is a trichotomy at the base all grasses except those of Anomochlooideae and Pharo- of the family, with these two genera and a clade consisting ideae, followed by a deletion within Ehrhartoideae, along the of all other grasses emerging from one point. line leading to Oryza in the present analysis. The present Apart from those genera placed by the present analysis data also imply a second deletion of this region, in Nardus with other elements of Pooideae (i.e., the clade consisting of (Fig. 2, internode A and two internodes AЈ). The GPWG Brachyelytrum and its sister group), this analysis includes (2001) analysis included an ndhF sequence from Nardus, but only three genera that were not sampled by the GPWG it was not definitive in the region of this indel, so this is the (2001). The placement of these three genera by the present first report of a deletion event in this region in Nardus. analysis is consistent with their conventional taxonomic Among taxa that have been examined for the inclusion of placements (Saccharum L. in Panicoideae, and Bambusa portions of ndhF and ndhH within the IR region, ndhH ex- Schreb. and Guadua Kunth in Bambusoideae). tends into the IR region in nearly all, including Joinvillea, Following the divergence of elements of Anomochlooi- Ecdeiocolea, Anomochloa, Streptochaeta, and Pharus. In deae from the clade that includes all other grasses, Pharo- five taxa (Amphipogon R. Br., Gynerium, Molinia Schrank, ideae and are the next two groups to diverge, in Sporobolus R. Br., and Pleuropogon R. Br., the first four of succession, from the clade that includes all remaining grass- which are elements of the PACCAD clade), the 5Ј-terminus (Fig. 2). Of the remaining grasses, the PACCAD clade is of ndhH does not extend into the IR region. In light of the resolved as monophyletic, as is the BEP clade, and these distribution of states of this character, the occurrence of a two clades are resolved as sister groups. The present sam- portion of ndhH within the IR region is interpreted as a 342 Davis and Soreng ALISO

Fig. 2.—Consensus of six most-parsimonious trees resolved by cladistic analysis of nucleotide sequence variation in ndhF, ndhH, matK, rbcL, and structural characters of the plastid and nuclear genomes from 108 taxa (Table 1). For representatives of Poaceae (i.e., all taxa except Joinvillea and Ecdeiocolea), according to Clayton and Renvoize (1986; with some changes in spelling) is indicated by two- codes in upper case for subfamily (AR ϭ Arundinoideae; BA ϭ Bambusoideae; CE ϭ Centothecoideae; ϭ Chloridoideae; PA ϭ Panicoideae; PO ϭ Pooideae), two-letter codes in lower case for tribe (ad ϭ ; an ϭ Anomochloeae; ar ϭ ; as ϭ ; av ϭ Aveneae; ba ϭ ; bc ϭ Brachyelytreae; br ϭ Bromeae; ce ϭ ; di ϭ ; eh ϭ ; er ϭ ; ha ϭ Hainardieae; ly ϭ Lygeeae; me ϭ Meliceae; na ϭ Nardeae; ol ϭ ; or ϭ ; pa ϭ Parianeae; ph ϭ Phareae; pn ϭ ; po ϭ Poeae; sc ϭ Streptochaeteae; st ϭ ; tr ϭ Triticeae), and three-letter codes in lower case for subtribes of Aveneae (alo ϭ Alopecurinae; ave ϭ Aveninae; dut ϭ Duthieinae; pha ϭ Phalaridinae). Taxonomy adopted by the GPWG (2001) is indicated by labels at right, with some modifications (see text). Numbers above internodes are jackknife percentages. Circles to the right of 24 names depict presence (closed) or absence (open) of GBSSI intron 10. Character transformations of selected characters are indicated by arrows as follow: A ϭ 15 base pair (bp) insertion in ndhF; AЈϭ15 bp deletion in ndhF; B ϭ extension of a portion of ndhF from the Short Single Copy region into the Inverted Repeat region (three other occurrences are not mapped—see text). VOLUME 23 Phylogeny of Poaceae Subfamily Pooideae 343

Fig. 3.—Continuation of consensus tree depicted in Fig. 2. Codes and labels are as in Fig. 2. Transformations of selected characters are indicated by arrows as follow: C ϭ three-nucleotide inversion in ndhF; CЈϭreversal of three-nucleotide inversion in ndhF; D ϭ loss of GBSSI intron 10; E ϭ six-nucleotide inversion in ndhF. 344 Davis and Soreng ALISO plesiomorphy for the grasses, shared at least with Joinville- grasses, with Anomochloa placed as the next lineage to di- aceae and Ecdeiocoleaceae, and the 5Ј-end of ndhH is inter- verge from the group that includes all other grasses, as in preted as having migrated out of the IR region once within some but not all of the most-parsimonious trees resolved by Pooideae (in Pleuropogon) and at least three times within the present analysis. The vegetative and reproductive struc- the PACCAD clade, where one of the transformations may tures of Anomochloa and Streptochaeta are quite different be synapomorphic for Amphipogon and Molinia. in form (Judziewicz and Soderstrom 1989), and these taxa The 3Ј-end of ndhF extends into the IR region in all taxa warrant further attention. of the PACCAD clade that have been examined, as well as The PACCAD clade, a group consisting of six of the re- in Ehrharta Thunb., Olyra L., and . maining nine subfamilies, as resolved by the GPWG (2001) This gene is interpreted as having terminated within the SSC and other analyses, also is resolved by the present analysis region in the earliest grasses, with its 3Ј-terminus having (Fig. 2). The extension of the 3Ј-end of ndhF into the IR migrated into the IR region four times, with one of these region appears to be a synapomorphy of this clade (Fig. 2, occurrences being a synapomorphy of the PACCAD clade character transformation B). Gynerium, a taxon of uncertain (Fig. 2, internode B; other three occurrences not depicted). relationship, was resolved by the GPWG analysis as the sis- ter of Panicoideae, and was treated there as incertae sedis. GBSSI Intron 10 It has since been assigned formally to that subfamily (Sa´n- chez-Ken and Clark 2001), within the tribe Gynerieae, and The presence or absence of intron 10 of GBSSI has been is treated here as an element of Panicoideae (Fig. 2). The determined for 24 taxa in the sample (Fig. 2, 3). Of these placement of Gynerium by the present analysis, as sister of 24 taxa, 13 have the intron and 11 lack it. Loss of the intron all other Panicoideae, is consistent with the results of the is interpreted as a nonhomoplasious synapomorphy of a GPWG analysis, as is the placement of Centothecoideae clade within Pooideae (Fig. 3, internode D). (represented here by Chasmanthium Link) as sister of Pan- icoideae, with this grouping of two subfamilies placed as the DISCUSSION sister of a clade that includes all remaining elements of the Two of the four genes used in the analysis, ndhF and PACCAD clade. An alternative taxonomic treatment is pro- matK, together contribute more than two-thirds of the char- vided by Zuloaga et al. (2003), who recognize an expanded acters in the analysis. One of these two, ndhF, is substantially Panicoideae that includes both Centotheceae (i.e., Centothe- longer than each of the other three, while matK is the second coideae) and Gynerieae. longest. Also, matK is more variable than any of the other With respect to placement of the endpoints of ndhF and three, as measured by the percentage of aligned sites that ndhH, relative to the boundaries between the SSC and IR are cladistically informative. The other two genes, ndhH and regions of the plastid genome, the present analysis supports rbcL, are similar to each other in length and in the percent- a different interpretation than that of Ogihara et al. (2002). ages of characters that are cladistically informative, and to- Those authors regarded the pattern that they observed as gether they contribute about a third of the total number of evidence of a closer relationship between Oryza and Triti- informative characters in the data set. Homoplasy levels are cum than between either of these taxa and Zea (i.e., the pres- similar among three of the genes, while matK, the outlier, ence of a substantial portion of the 5Ј-end of ndhH in the has the highest consistency and retention indices of the four. IR region in Oryza and Triticum, as opposed to the presence Thus, matK is both more variable and less homoplasious of only one nucleotide of ndhH in the IR region in Zea; and than the other three genes. the presence of the 3Ј-end of ndhF in the IR region in Zea, Relationships resolved by the present analysis should be as opposed to termination of ndhF within the SSC region in regarded as tentative, and for this reason we do not discuss Oryza and Triticum). The present analysis does support a all noteworthy aspects of the phylogenetic results, particu- closer phylogenetic relationship between Oryza and Triticum larly outside subfamily Pooideae. However, some aspects of than between either of these taxa and Zea, as have most these results deserve comment. The present analysis is con- previous phylogenetic analyses that have addressed this mat- sistent with the GPWG (2001) analysis in placing represen- ter (e.g., GPWG 2001 and several papers cited therein). In- tatives of Anomochlooideae, Pharoideae, and Puelioideae as deed, any analysis that resolves a BEP clade, with Oryza in the first lineages to diverge within the grass family from a Ehrhartoideae and Triticum in Pooideae, supports a closer large clade that includes the remaining nine subfamilies (Fig. relationship between these two taxa than between either of 2), and that is marked by the 15-nucleotide insertion in ndhF. them and Zea. However, placement of the endpoints of ndhF Within this clade, the length of ndhF in this region is re- and ndhH, relative to the boundaries between the major plas- stored to its original (i.e., shorter) state in two of the sampled tid genomic regions, does not in itself provide evidence in taxa, Ehrharta and Nardus, by what appear to be separate support of this relationship. As demonstrated here, the states 15-nucleotide deletion events (Fig. 2, character transforma- of these characters, as observed in Oryza and Triticum, also tion AЈ. Despite the general similarity of the present results occur in the early diverging lineages within the grasses, as to those of the GPWG analysis, with respect to the early well as outside the family, and thus are best interpreted as diverging grass lineages, the absence in some trees of a clade plesiomorphies of the grasses. What these features do pro- consisting of Anomochloa and Streptochaeta suggests that vide, in terms of phylogenetic evidence, is support for mono- these two anomalous grasses may not be closest relatives. phyly of the PACCAD clade, as already indicated. Extension The placement of these two genera as separately diverging of the 3Ј-end of ndhF into the IR region is observed in all lineages is consistent with the results of Hilu et al. (1999), taxa of the PACCAD clade that have been examined, as well whose analysis placed Streptochaeta as the sister of all other as in three other isolated taxa. These occurrences, in the VOLUME 23 Phylogeny of Poaceae Subfamily Pooideae 345 context of the relationships resolved by this analysis, are best GBSSI intron 10 as a synapomorphy of this clade (Fig. 3, explained as having arisen by four parallel migrations of the character transformation D). This clade has high jackknife 3Ј-end of ndhF into the IR region. One of these occurrences support (100%) with the intron character either included or marks the PACCAD clade, and the other three occurrences, excluded from the analysis. Milium is one of the 11 taxa for once within each subfamily of the BEP clade, are interpreted which the absence of this intron has been observed, and this as autapomorphies. congruence between a marker in the nuclear genome and the Like the analysis by the GPWG (2001), and various ear- plastid-encoded genes that determine the structure of the tree lier analyses, the present study resolves the BEP clade. With- suggests that this does belong within Poeae s.l. in this clade, the present analysis resolves all three subfam- The present analysis also agrees with that of Soreng and ilies as monophyletic (Fig. 2), as have several earlier anal- Davis (2000) in placing a clade consisting of Bromeae and yses. However, the present analysis places Bambusoideae Triticeae as sister of Poeae s.l. (Fig. 3), thus forming a group and Pooideae as sister taxa, with 97% jackknife support, that we designate the BT/P clade (93% jackknife support). with Ehrhartoideae placed as the sister of the clade that in- It should be noted, however, that Bromeae are not mono- cludes these two. In contrast, the GPWG (2001) analysis, phyletic in the present analysis, because of the placement of like that of Clark et al. (1995) and others, placed Ehrharto- Littledalea Hemsl. as sister of Bromus L. plus Triticeae. A ideae and Bambusoideae as sisters, with Pooideae the sister similar result is reported by Saarela et al. (2007), whose of the clade that consisted of those two subfamilies. Alter- sampling within Bromeae is broader than that of the present native phylogenetic structures for these three subfamilies analysis; Littledalea was not sampled by Soreng and Davis have been resolved by other analyses, including those of (2000). A critical difference between the results of the pre- various subsets of the GPWG data matrix, and in some anal- sent analysis and those of Soreng and Davis (2000) lies in yses there is no BEP clade (e.g., Soreng and Davis 1998; the placement of Brachypodieae (represented by Brachypo- Hilu et al. 1999). Even when the BEP clade is absent, the dium), rather than a clade consisting of Brachypodieae plus PACCAD clade generally is resolved, often with Pooideae Meliceae, as sister of the BT/P clade. The present analysis placed as its sister, with Bambusoideae and Ehrhartoideae identifies a three-nucleotide inversion in ndhF as a synapo- then situated either as sister taxa or as successively diverging morphy of the clade that consists of Brachypodieae and its lineages nearby. A complete review of the various arrange- sister group, the BT/P clade (Fig. 3, character transformation ments that have been resolved among these taxa by other C). This inversion has not been observed in any other spe- data sets is beyond the scope of this paper, and we simply cies in the sample, including the various representatives of note that relationships among Bambusoideae, Ehrhartoideae, Meliceae. This character is homoplasious, since it is reversed Pooideae, and the PACCAD clade are not yet firmly estab- to the uninverted state in Hordeum (Fig. 3, transformation lished, and refer readers to the trees depicted in Appendix CЈ), but the placement of Meliceae as sister of Brachypo- III of the GPWG (2001) analysis, and to the various contri- dieae would require an additional homoplasious transfor- butions cited in that paper. mation. Brachypodieae also were placed as the sister of Bro- The most comprehensive previous analysis of phyloge- meae, Triticeae, and Poeae s.l., with Meliceae placed outside netic relationships within Pooideae was that of Soreng and this group, by Hilu et al. (1999). Davis (2000), which was based on morphological characters Several other differences within Pooideae, with respect to and restriction sites from the plastid genome. There the fol- relationships among tribes and isolated genera, also exist be- lowing set of relationship among conventionally delimited tween the trees resolved by the present analysis and those tribes and isolated genera was resolved within the subfamily: of Soreng and Davis (2000). The two analyses agree in plac- (Brachyelytreae ((Lygeeae Nardeae) ((Anisopogon R. Br. ing Brachyelytreae as the sister of all other elements of Pooi- (Stipeae including Ampelodesmos Link)) (Diarrheneae ((Bra- deae, and in placing Lygeeae and Nardeae as sisters of each chypodieae Meliceae) ((Bromeae Triticeae) Poeae s.l.)))))). other, with this small clade the sister of one that consists of Although Brachyelytrum was used as the outgroup for the all remaining elements of the subfamily (Fig. 2). The two published tree in that analysis, preliminary analyses utilizing analyses also agree in placing Ampelodesmos among genera taxa outside Pooideae established the sister group relation- conventionally assigned to Stipeae (Fig. 2). However, the ship between this genus and other members of the subfamily. analysis of Soreng and Davis (2000) places Anisopogon as The present analysis includes 77 representatives of Pooideae, the sister of this overall group, which itself is the sister of a 24 fewer taxa overall than the 101 representatives sampled clade that includes all representatives of Pooideae except by Soreng and Davis (2000), but with multiple species sam- Brachyelytreae, Lygeeae, and Nardeae. The present analysis pled within only three genera in the present study, as op- differs in placing Anisopogon with three additional disparate posed to 18 genera sampled by two or three species each in genera, Duthiea Hack., Sinochasea Keng, and Phaenosper- the earlier analysis. Each analysis includes some taxa that ma Munro ex Benth. (Fig. 3). This group of four genera the other does not, so all relationships resolved by one can- diverges just after Stipeae (including Ampelodesmos) from not be compared to those in the other. However, there is the clade that includes the BT/P clade. We tentatively refer sufficient overlap in taxon sampling to facilitate comparison to this group of four genera (plus other elements of subtribe of the results. In each case, taxa of the traditionally recog- Duthieinae sensu Clayton and Renvoize 1986) as Phaeno- nized tribes Aveneae, Hainardieae, and Poeae (i.e., Poeae spermateae. Of the taxa in this group, only Anisopogon was s.s.), plus Milium L., which was included in Stipeae by Clay- sampled by Soreng and Davis (2000), and the present anal- ton and Renvoize (1986), are placed within a clade that we ysis is the first report of a phylogenetic placement of Duthiea continue to designate as the tribe Poeae s.l. In the present and Sinochasea on the basis of DNA sequence variation. The analysis we identify the apparently nonhomoplasious loss of GPWG (2001) resolved a clade consisting of Phaenosperma 346 Davis and Soreng ALISO and Anisopogon, but did not sample the other two genera and Davis (2000) resolve two major groups of taxa tradi- that are placed in the group by the present analysis. This set tionally assigned to Aveneae, which can be recognized as of genera and their presumed relatives have been difficult to subtribes (though not all of the following genera were sam- classify on the basis of traditional characters. Anisopogon pled in both studies): Aveninae (Avena, Gaudinia, Helicto- was placed in Arundinoideae by Clayton and Renvoize trichon, Rostraria, Trisetum); and Agrostidinae (Agrostis L., (1986) and Watson and Dallwitz (1992). Duthiea and Sino- Host, Calamagrostis Adans., Gastridium P. chasea (the latter as a synonym of Pseudodanthonia Bor & Beauv., Polypogon Desf., Triplachne Link). The Aveninae C. E. Hubb.) were placed, with Metcalfia Conert and Ste- clade is marked by the nonhomoplasious six-nucleotide in- phanachne Keng, in Pooideae tribe Aveneae subtribe Du- version in ndhF described above (Fig. 3, character transfor- thieinae by Clayton and Renvoize, while Watson and Dall- mation E). Other isolated elements are placed among these witz expressed their doubts by placing these either in Arun- clades, with positions of some of the genera differing be- dinoideae tribe Danthonieae or Pooideae tribe Aveneae, in tween the studies. Subtribe Phalaridinae, represented by An- both cases annotated with question marks. Phaenosperma thoxanthum L. in the present study, was resolved by Soreng was placed in Bambusoideae, in its own tribe, Phaenosper- and Davis (2000) as sister of a clade that included elements mateae, by Clayton and Renvoize, and in the supertribe Ory- of Briza s.l. plus Agrostidinae, but it is placed as sister of zodae by Watson and Dallwitz. Aveninae in the present analysis. The six-nucleotide inver- The two remaining tribes of Pooideae, Diarrheneae and sion is absent in Anthoxanthum, so this character does not Meliceae, also are placed differently by the present analysis favor either of these placements over the other. It is of in- and that of Soreng and Davis (2000). In the present analysis, terest that Briza s.l. is non-monophyletic in both studies, and Meliceae and Diarrheneae diverge in succession from the that in the present study Echinopogon P. Beauv. (endemic to large group that includes Brachypodieae and the BT/P clade ) is placed, within a pectinate structure, between (Fig. 2, 3), but the analysis of Soreng and Davis (2000) Briza s.s. from and other elements of Briza s.l. from placed Diarrheneae as the first of these elements to diverge (including Calotheca and Chascolytrum, after Stipeae plus Anisopogon, with Meliceae plus Brachy- which were also associated with Poidium Nees in the earlier podieae the next lineage to diverge. analysis). Also, Amphibromus Nees was the sister of all oth- In light of these various similarities and differences, the er elements of subclade 1 in the analysis of Soreng and Da- picture that emerges is one in which Brachyelytreae and a vis (2000), but in the present analysis it is joined in this clade consisting of Lygeeae plus Nardeae appear to have position by Torreyochloa (i.e., these two taxa constitute a diverged in succession from the lineage that includes all oth- clade that is placed as sister of the rest of subclade 1). This er members of Pooideae. Within the latter group are five clade of two genera also was resolved by Soreng and Davis smaller groups, these being Diarrheneae, Meliceae, Phae- (2000) in analyses of only the restriction site data, and these nospermateae (should it continue to be resolved), Stipeae wetland genera are provisionally recognizable as constituting (including Amplelodesmos), and a clade consisting of Bra- the small subtribe Torreyochloinae. chypodieae and the BT/P clade, which represents the major Within subclade 2, the present study and that of Soreng radiation within the subfamily. Relationships among these and Davis (2000) resolve some of the same groups of taxa: five clades still are unclear. (A) intermixed elements of subtribes Alopecurinae s.s. (Al- We turn now to Poeae s.l., which corresponds closely to opecurus L., Beckmannia Host, Phleum L.), Poineae (Arc- Poeae s.l. as resolved by Soreng and Davis (2000), though tagrostis Griseb., Bellardiochloa Chiov., Dupontia R. Br., the taxon sampling varies between that analysis and the pre- Poa L.), and Miliinae (Milium); (B) subtribes Dactylidinae sent one. Within Poeae s.l., the present analysis resolves two ( L., which is placed here with restriction site data principal clades, labeled as Poeae s.l. subclades 1 and 2 in by Soreng and Davis 2000, but not when combined with Fig. 3. To avoid excessive repetition, we refer to these clades morphology), Cynosurinae (Cynosurus L.), and Parapholi- through the remainder of the discussion as subclades 1 and inae (Parapholis C. E. Hubb.), with Sphenopus Trin., Cu- 2. Subclade 1 consists mostly of elements of Clayton and tandia Willk., and Desmazeria Dumort.; (C) Loliinae (Fes- Renvoize’s (1986) tribe Aveneae (minus subtribe Duthiei- tuca L., Leucopoa Griseb., Lolium L., Vulpia C. C. Gmel.) nae), but it also includes four taxa that those authors as- with Dielsiochloa Pilg.; (D) Puccinelliinae (Catabrosa P. signed to Poeae s.s. (Torreyochloa G. L. Church, Briza L. Beauv., Puccinellia Parl., Sclerochloa P. Beauv.); and (E) s.s., Calotheca Desv., and Chascolytrum Desv., the latter two Sesleriinae (incl. Miborinae; Sesleria Scop., Mibora Adans.). of which they included in Briza s.l.). Subclade 2 includes A fairly close relationship was detected between clades B most of the sampled taxa from Clayton and Renvoize’s tribe and C in both analyses, but the sister of clade D was clade Poeae, plus tribe Hainardieae, Milium (which they assigned C in the analysis of Soreng and Davis (2000), while the to Stipeae), and several groups of genera that they assigned sister of clade D is clade A in the present analysis, with 73% to Aveneae. jackknife support (Fig. 3). Sesleria was resolved as the sister A reconsideration of relationships in Pooideae, on the ba- of clade A in the previous analysis, but in the present anal- sis of morphologically defined groupings and ongoing phy- ysis, as part of group E (i.e., along with Mibora), it is placed logenetic studies, is reflected in the revised classification of with 98% jackknife support as sister to a clade that includes the subfamily proposed by Soreng et al. (2003), which is clades B and C, plus a clade consisting of Holcus L. (Hol- updated at: http://mobot.mobot.org/W3T/Search/nwgclass. cinae) and Vahlodea Fr., and another, consisting of Des- . Suprageneric taxa in the following discussion reflect champsia P. Beauv. and Molineriella Rouy (Fig. 3). groups that are recognized within this developing system. One of the most complex situations within Poeae s.l. in- Within subclade 1, the present analysis and that of Soreng volves Airinae (Aira L., Deschampsia s.l. [including Ave- VOLUME 23 Phylogeny of Poaceae Subfamily Pooideae 347 nella (Bluff & Fingerh.) Drejer and Vahlodea], Holcinae WILLIAMS, P. A. GADEK, C. J. QUINN, L. E. EGUIARTE, E. GOLEN- [Holcus], and Molineriella). The analysis of Soreng and Da- BERG, G. H. LEARN,JR., S. C. H. BARRETT, S. DAYANANDAN, AND vis (2000) also placed (Dumort.) Dumort. in sub- . A. ALBERT. 1993. Phylogenetics of seed plants: an analysis of clade 2, among genera of Airinae and Holcinae. In light of nucleotide sequences from the plastid gene rbcL. Ann. Bot. Gard. 40: 528–580. the discrepancy in the placement of CLARK, L. G., M. KOBAYASHI, S. MATHEWS, R. E. SPANGLER, AND E. between the two studies (it was placed in subclade 1 by the A. KELLOGG. 2000. The Puelioideae, a new subfamily of Poaceae. analysis of Soreng and Davis 2000), we re-examined the Syst. Bot. 25: 181–187. seed sample sent as that species by the USDA Plant Intro- , W. ZHANG, AND J. F. WENDEL. 1995. A phylogeny of the duction Station and used by Soreng and Davis (2000), and grass family (Poaceae) based on ndhF sequence data. Syst. Bot. have determined that accession as a species of Agrostis s.s. 20: 436–460. (plants grown from this seed sample have not flowered). CLAYTON, W. D., AND S. A. RENVOIZE. 1986. Genera graminum: Thus, all verified accessions of Deschampsia and its segre- grasses of the world. Kew Bull., Addit. Ser. 13: 1–389. gates are placed by molecular characters within subclade 2. DAHLGREN, R. M. T., H. T. CLIFFORD, AND P. F. YEO. 1985. The With this matter set aside, Airinae still appear to be non- families of . Springer-Verlag, , Germany. 520 p. monophyletic on the basis of variation patterns in their chlo- DAVIS, J. I., AND R. J. SORENG. 1993. Phylogenetic structure in the roplast DNA. The sampling of elements that have been rec- grass family (Poaceae) as inferred from chloroplast DNA restric- ognized within Aveneae subtribe Airinae is more complete tion site variation. Amer. J. Bot. 80: 1444–1454. in the present analysis than in that of Soreng and Davis DOYLE, J. J., J. I. DAVIS, R. J. SORENG, D. GARVIN, AND M. J. AN- (2000), and these taxa are scattered among three lineages DERSON. 1992. Chloroplast DNA inversions and the origin of the within subclade 2. Deschampsia s.l. (including Deschampsia grass family (Poaceae). Proc. Natl. Acad. Sci. U.S.A. 89: 7722– s.s., Avenella, and Vahlodea) is not monophyletic, and each 7726. of three representatives of this genus is placed in a small DUVALL, M. R., J. I DAVIS, L. G. CLARK, J. D. NOLL, D. H. GOLD- clade with another genus (Molineriella, Aira, and Holcus, MAN, AND J. G. SA´ NCHEZ-KEN. 2007. Phylogeny of the grasses respectively) in disparate locations within subclade 2, where (Poaceae) revisited, pp. 237–247. In J. T. Columbus, E. A. Friar, J. M. Porter, L. M. Prince, and M. G. Simpson [eds.], Monocots: they are removed from each other and from the other ele- comparative biology and evolution—Poales. Rancho Santa Ana ments of Aveneae s.l. in subclade 1 (Fig. 3). Botanic Garden, Claremont, , USA. The present analysis confirms previous observations that FARRIS, J. S., V. A. ALBERT, M. KA¨ LLERSJO¨ , D. LIPSCOMB, AND A. G. the traditionally recognized tribes Poeae s.s. and Aveneae KLUGE. 1996. Parsimony jackknifing outperforms neighbor-join- are phylogenetically intermixed, at least as determined by ing. Cladistics 12: 99–124. chloroplast DNA characters. Some of the cases described FRUGOLI, J. A., M. A. MCPEEK, T. L. THOMAS, AND C. R. MCCLUNG. above may be attributable to error in the structures of these 1998. Intron loss and gain during evolution of the catalase gene cladograms, relative to the actual history of diversification family in angiosperms. Genetics 149: 355–365. of the plastid genomes, while others may signify homoplas- GOLOBOFF, P. A. 1993. NONA vers. 1.6. Tucuma´n, . Pub- tic changes in the morphological characters that underlie tra- lished by the author. Available at: http://www.cladistics.org/ education.html (Jun 2005). ditional classifications. Reticulation may be a third cause of GRASS PHYLOGENY WORKING GROUP [GPWG]. 2001. Phylogeny and these unexpected results. subfamilial classification of the grasses (Poaceae). Ann. Missouri Bot. Gard. 88: 373–457. ACKNOWLEDGMENTS HILU, K. W., L. A. ALICE, AND H. LIANG. 1999. Phylogeny of Po- We thank several colleagues for providing plant materials aceae inferred from matK sequences. Ann. Missouri Bot. Gard. and DNA isolations used in this work, and Pilar Catala´n for 86: 835–851. HIRATSUKA, J., H. SHIMADA, R. WHITTIER, T. ISHIBASHI, M. SAKA- kindly inviting us to present this work at the symposium on MOTO, M. MORI, C. KONDO, . HONJI, C. R. SUN, B. Y. MENG, Y. Systematics of Pooideae at Monocots III and Grasses IV. . LI, A. KANNO, Y. NISHIZAWA, A. IHRAI, K. SHINOZAKI, AND M. We also thank the Smithsonian Institution (SI) Biodiversity SUGIURA. 1989. The complete sequence of the rice (Oryza sativa) Surveys and Inventories Programs for support to RJS for chloroplast genome: intermolecular recombination between dis- plant collecting in , the SI Research Opportunities tinct tRNA genes accounts for a major plastid DNA inversion Fund for support to RJS for collecting in Australia, and the during the evolution of the . Molec. Gen. Genet. 217: 185– National Science Foundation for support of this project 194. (grant DEB-0318686). HOLMGREN, P. K., N. H. HOLMGREN, AND L. C. BARNETT. 1990. Index herbariorum. Part I: the herbaria of the world. Ed. 8. New York LITERATURE CITED Botanical Garden, New York, USA. 693 p. Available at: http:// sciweb.nybg.org/science2/IndexHerbariorum.asp (Jun 2005). ASMUSSEN, C. B., AND M. W. CHASE. 2001. Coding and noncoding HSIAO, C., S. W. L. JACOBS, N. J. CHATTERTON, AND K. H. ASAY. plastid DNA in palm systematics. Amer. J. Bot. 88: 1103–1117. 1999. A molecular phylogeny of the grass family (Poaceae) based BREMER, K. 2002. Gondwanan evolution of the grass alliance of on the sequences of nuclear ribosomal DNA (ITS). Austral. Syst. families. Evolution 56: 1374–1387. Bot. 11: 667–688. CHASE, M. W., D. E. SOLTIS, R. G. OLMSTEAD, D. MORGAN, D. H. INGRAM, A. L., AND J. J. DOYLE. 2003. The origin and evolution of LES, B. D. MISHLER, M. R. DUVALL, R. A. PRICE, H. G. HILLS, Y.- tef (Poaceae) and related polyploids: evidence from L. QIU, K. A. KRON, J. H. RETTIG, E. CONTI, J. D. PALMER, J. R. nuclear waxy and plastid rps16. Amer. J. Bot. 90: 116–122. MANHART, K. J. SYTSMA, H. J. MICHAELS, W. J. KRESS, K. J. KA- ITCHODA, N., S. NISHIZAWA, H. NAGANO, T. KUBO, AND T. MIKAMI. ROL, W. D. CLARK, M. HEDRE´ N, B. S. GAUT, R. K. JANSEN, K.-J. 2002. The sugar beet mitochondrial nad4 gene: an intron loss and KIM, C. F. WIMPEE, J. F. SMITH, G. R. FURNIER, S. H. STRAUSS, Q.- its phylogenetic implication in the Caryophyllales. Theor. Appl. Y. XIANG, G. M. PLUNKETT, P. S. SOLTIS, S. M. SWENSEN, S. E. Genet. 104: 209–213. 348 Davis and Soreng ALISO

JUDZIEWICZ, E. J., AND T. R. SODERSTROM. 1989. Morphological, an- PLUNKETT, G. M., AND S. R. DOWNIE. 2000. Expansion and contrac- atomical, and taxonomic studies in Anomochloa and Streptochaeta tion of the chloroplast inverted repeat in Apiaceae subfamily (Poaceae: Bambusoideae). Smithsonian Contr. Bot. 68: 1–52. Apioideae. Syst. Bot. 25: 648–667. KATAYAMA, H., AND Y. OGIHARA. 1996. Phylogenetic affinities of the SAARELA, J. M., P. M. PETERSON, R. M. KEANE, J. CAYOUETTE, AND grasses to other monocots as revealed by molecular analysis of S. W. GRAHAM. 2007. Molecular phylogenetics of Bromus (Po- chloroplast DNA. Curr. Genet. 29: 572–581. aceae: Pooideae) based on chloroplast and nuclear DNA sequence KELLOGG, E. A., AND C. S. CAMPBELL. 1987. Phylogenetic analyses data, pp. 450–467. In J. T. Columbus, E. A. Friar, J. M. Porter, L. of the Gramineae, pp. 310–322. In T. R. Soderstrom, K. W. Hilu, M. Prince, and M. G. Simpson [eds.], Monocots: comparative bi- C. S. Campbell, and M. E. Barkworth [eds.], Grass systematics ology and evolution—Poales. Rancho Santa Ana Botanic Garden, and evolution. Smithsonian Institution Press, Washington, D.C., Claremont, California, USA. USA. SA´ NCHEZ-KEN, J. G., AND L. G. CLARK. 2001. Gynerieae, a new MASON-GAMER, R. J. 2001. Origin of North American Elymus (Po- Neotropical tribe of grasses (Poaceae). Novon 11: 350–352. aceae: Triticeae) allotetraploids based on granule-bound starch SORENG, R. J., AND J. I. DAVIS. 1998. Phylogenetics and character synthase gene sequences. Syst. Bot. 26: 757–768. evolution in the grass family (Poaceae): simultaneous analysis of , C. F. WEIL, AND E. A. KELLOGG. 1998. Granule-bound starch morphological and chloroplast DNA restriction site character sets. synthase: structure, function, and phylogenetic utility. Molec. Biol. Bot. Rev. (Lancaster) 64: 1–85. Evol. 15: 1658–1673. , AND . 2000. Phylogenetic structure in Poaceae sub- MATHEWS, S., R. E. SPANGLER, R. J. MASON-GAMER, AND E. A. KEL- family Pooideae as inferred from molecular and morphological LOGG. 2002. Phylogeny of Andropogoneae inferred from phyto- characters: misclassification vs. reticulation, pp. 61–74. In S. W. chrome B, GBSSI, and ndhF. Int. J. Pl. Sci. 163: 441–450. L. Jacobs and J. E. Everett [eds.], Grasses: systematics and evo- , R. C. TSAI, AND E. A. KELLOGG. 2000. Phylogenetic struc- lution. CSIRO Publishing, Collingwood, , Australia. ture in the grass family (Poaceae): evidence from the nuclear gene , P. M. PETERSON, G. DAVIDSE, E. J. JUDZIEWICZ, F. O. ZU- phytochrome B. Amer. J. Bot. 87: 96–107. LOAGA, T. S. FILGUEIRAS, AND O. MORRONE (editors). 2003. Cata- MICHELANGELI, F. A., J. I DAVIS, AND D. W. STEVENSON. 2003. Phy- logue of New World grasses (Poaceae): IV. Subfamily Pooideae. logenetic relationships among Poaceae and related families as in- Contr. U.S. Natl. Herb. 48: 1–730. ferred from morphology, inversions in the plastid genome, and STEBBINS, G. L., AND B. CRAMPTON. 1961. A suggested revision of sequence data from the mitochondrial and plastid genomes. Amer. the grass genera of temperate . Recent Advances in J. Bot. 90: 93–106. Botany 1: 133–145. NIXON, K. C. 2002. WinClada vers. 1.00.08. Ithaca, New York, USA. WALLACE, R. S., AND J. H. COTA. 1996. An intron loss in the chlo- Distributed by the author. Available at: http://www.cladistics.org/ roplast gene rpoC1 supports a monophyletic origin for the sub- education.html (Jun 2005). family Cactoideae of the Cactaceae. Curr. Genet. 29: 275–281. OGIHARA, Y., K. ISONO, T. KOJIMA, A. ENDO, M. HANAOKA, T. SHIINA, WATSON, L., AND M. J. DALLWITZ. 1992 onwards. The families of T. TERACHI, S. UTSUGI, M. MURATA, N. MORI, S. TAKUMI, K. IKEO, flowering plants: descriptions, illustrations, identification, and in- T. GOJOBORI, R. MURAI, K. MURAI, Y. MATSUOKA, Y. OHNISHI, H. formation retrieval, Vers. 14 Dec 2000. http://biodiversity.uno. TAJIRI, AND K. TSUNEWAKI. 2002. Structural features of a edu/delta/angio [Now at: http://delta-intkey.com Mar 2006.] plastome as revealed by complete sequencing of chloroplast DNA. ZHANG, W. 2000. Phylogeny of the grass family (Poaceae) from Molec. Genet. Genomics 266: 740–746. rpl16 intron sequence data. Molec. Phylogen. Evol. 15: 135–146. OLMSTEAD, R. G., AND J. A. SWEERE. 1994. Combining data in phy- ZULOAGA, F. O., O. MORRONE, G. DAVIDSE, T. S. FILGUEIRAS, P. M. logenetic systematics: an empirical approach using three molec- PETERSON, R. J. SORENG, AND E. J. JUDZIEWICZ (editors). 2003. ular data sets in the Solanaceae. Syst. Biol. 43: 467–481. Catalogue of New World grasses (Poaceae): III. Subfamilies Pan- PALMER, J. D. 1983. Chloroplast DNA exists in two orientations. icoideae, Aristidoideae, Arundinoideae, and Danthonioideae. Nature 301: 92–93. Contr. U.S. Natl. Herb. 46: 1–662.